Work at SourceForge, help us to make it a better place! We have an immediate need for a Support Technician in our San Francisco or Denver office.

Close

Graph Mining papers (Graph Kernel and Maximum Common Subst.)

Two actual papers are published in the 'QSAR & Comb. Chem.' journal http://dx.doi.org/10.1002/qsar.200510135 and http://dx.doi.org/10.1002/qsar.200510009

The Maximum Common Substructure (MCS) method works uses single atom properties, which can be also numeric.

The Optimal Assignment (Graph) Kernel works on multiple chemical atom and bond labels and is an extremely fast chemical similarity measure. Due to the fact that it is a positive definite matrix, this measure can be used for any SVM based learning method, like classification, regression, and clustering.

Here are the related BibTeX entries generated with JabRef http://jabref.sourceforge.net.
----

@ARTICLE{fwz05c,
author = {H. Fr{\"{o}}hlich and J. K. Wegner and A. Zell},
title = {{K}ernel {F}unctions for {A}ttributed {M}olecular {G}raphs -- {A}
{N}ew {S}imilarity {B}ased {A}pproach {T}o {ADME} {P}rediction in
{C}lassification and {R}egression},
journal = {QSAR Comb. Sci.},
year = {2005},
pages = {accepted},
doi = {10.1002/qsar.200510135},
owner = {wegner@users.sf.net},
url ={ http://dx.doi.org/10.1002/qsar.200510135 }
}

@ARTICLE{wfmz05,
author = {J. K. Wegner and H. Fr{\"{o}}hlich and H. Mielenz and A. Zell},
title = {{D}ata and {G}raph {M}ining in {C}hemical {S}pace for {ADME} and
{A}ctivity {D}ata {S}ets},
journal = {QSAR Comb. Sci.},
year = {2005},
pages = {accepted},
doi = {10.1002/qsar.200510009},
owner = {wegner@users.sf.net},
url ={ http://dx.doi.org/10.1002/qsar.200510009 }
}

Posted by Joerg Kurt Wegner 2006-01-30